China's yongyuan hesheng raises $7.5M to mass-produce biological data for AI
What's the deal? Chinese startup yongyuan hesheng has raised US$6.88M (around $7.5 million) in an angel round. Shunwei Capital and Yunmu CapitalDealroom has a profile for this one. Try Dealroom → backed the round, which closed in November 2026.
What does the company do? Founded in 2022 and based in Shenzhen, yongyuan hesheng calls itself a "data foundry platform" for AI and biology. It builds DNA sequences, runs functional experiments, and packages the results into AI-ready sequence-to-function datasets for research institutions, drugmakers, and model teams.
The core bet: The company's entry point is long-fragment DNA synthesis, which it positions as the infrastructure for training AI biology models. Founder and chief executive officer Yang Zongying argues the real bottleneck for AI in biology is not compute or algorithms but the scarcity of biological training data — a shortage he traces to the high cost of synthesising DNA.
How it works: Yang calls the method "DNA movable-type printing." Instead of assembling single nucleotides one at a time, the company draws from a prebuilt library of double-stranded modules that can be reused like movable type. It says this cuts a 1,000-base synthesis from around 999 reactions to roughly 99, while raising single-base fidelity to 99.95%.
The numbers it claims: yongyuan hesheng says its per-base synthesis cost has fallen below US$0.01, about a tenth of typical commercial levels, and that it delivers 10kb DNA in under seven days versus the usual 21 to 35. The technique is stable in the lab; the company is now working on factory-scale production.
What the money is for? Proceeds will fund high-value datasets, expand the company's proprietary DNA synthesis capacity, and build out its data foundry infrastructure. It also runs a rapid protein characterisation system it says can process more than 2,000 samples a week.
What's the endgame? Beyond cheaper DNA, the company pairs each sequence with measured function — for antibodies, that means binding strength against a target — and records full metadata so models can filter noise. It sells both off-the-shelf and custom datasets, and has worked with a Fortune 500 firm on mRNA UTR design. Yang expects revenue "in the tens of millions" of CNY by 2027.
What could go wrong? This remains an early technical path, and the round itself does not prove the data's quality or commercial demand. Factory-scale output, synthesis costs, and whether the data actually improves model performance are all still unproven.
The signal: Investors are betting on the upstream layer — scalable manufacturing of biological data — as a foundation for AI in the life sciences. Recent deals show continued appetite for DNA synthesis and automation, but synthetic-biology backers increasingly want customers and revenue, not just platform concepts.
Read more: 36kr.com
Image credit: Generated with Gemini
